Object recognition and categorization: some lessons from psychophysics, neurobiology and computer vision

Shimon Edelman
Department of Psychology
Cornell University
Ithaca, NY 14853, USA
http://kybele.psych.cornell.edu/$\sim$edelman

Much useful information about a visual object can be obtained by computing its similarities to a small number of reference shapes or prototypes, which, in turn, can be represented by their view spaces, interpolated from a handful of exemplar views. Such low-dimensional, hence computationally tractable, view-based representations support both the recognition of familiar shapes and the categorization of novel ones [1]. Apart from categorization, they can also be used in a variety of other tasks involving novel objects: viewpoint-insensitive recognition, recovery of a canonical view, and estimation of pose or of arbitrary novel views [2]. Predictions generated by this computational model concerning the cortical physiology of object representation in primates have been borne out by experiments (e.g., [3,4,5]). Moreover, its limitations vis-à-vis dealing with progressive shape change and with image translation (as well as other stimulus manipulations) resemble those of human subjects [6,7]. However, the absolute level of performance of the implemented system that had been based on this approach [8] fell short of the human standard. In this talk, I shall discuss possible approaches to closing this performance gap while keeping the model computationally feasible and biologically relevant.

Bibliography

1

S. Edelman.
Representation is representation of similarity.
Behavioral and Brain Sciences, 21:449-498, 1998.

2

S. Edelman and S. Duvdevani-Bar.
Similarity-based viewspace interpolation and the categorization of 3D objects.
In Proc. Similarity and Categorization Workshop, pages 75-81, Dept. of AI, University of Edinburgh, 1997.

3

N. K. Logothetis, J. Pauls, and T. Poggio.
Shape recognition in the inferior temporal cortex of monkeys.
Current Biology, 5:552-563, 1995.

4

D. J. Freedman, M. Riesenhuber, T. Poggio, and E. K. Miller.
Categorical representation of visual stimuli in the primate prefrontal cortex.
Science, 291:312-316, 2001.

5

H. Op de Beeck, J. Wagemans, and R. Vogels.
Inferotemporal neurons represent low-dimensional configurations of parameterized shapes.
Nature Neuroscience, 4:1244-1252, 2001.

6

S. Edelman.
Representation and recognition in vision.
MIT Press, Cambridge, MA, 1999.

7

M. Dill and S. Edelman.
Imperfect invariance to object translation in the discrimination of complex shapes.
Perception, 30:707-724, 2001.

8

S. Duvdevani-Bar and S. Edelman.
Visual recognition and categorization on the basis of similarities to multiple class prototypes.
Intl. J. Computer Vision, 33:201-228, 1999.

About this document ...

Object recognition and categorization: some lessons from psychophysics, neurobiology and computer vision

This document was generated using the LaTeX2HTML translator Version 2002 (1.62)

Copyright © 1993, 1994, 1995, 1996, Nikos Drakos, Computer Based Learning Unit, University of Leeds.
Copyright © 1997, 1998, 1999, Ross Moore, Mathematics Department, Macquarie University, Sydney.

The command line arguments were:
latex2html -split 0 -nonavigation abstract.tex

The translation was initiated by Shimon Edelman on 2004-04-08